Gossip: Identifying Central Individuals in a Social Network
Last registered on November 27, 2016

Pre-Trial

Trial Information
General Information
Title
Gossip: Identifying Central Individuals in a Social Network
RCT ID
AEARCTR-0001770
Initial registration date
November 27, 2016
Last updated
November 27, 2016 10:53 PM EST
Location(s)
Region
Primary Investigator
Affiliation
MIT
Other Primary Investigator(s)
PI Affiliation
Stanford University
PI Affiliation
Stanford University
PI Affiliation
MIT
Additional Trial Information
Status
Completed
Start date
2015-01-29
End date
2015-07-15
Secondary IDs
Abstract
Is it possible, simply by asking a few members of a community, to identify individuals who are best placed to diffuse information? A simple model of diffusion shows how boundedly rational individuals can, just by tracking gossip about people, identify those who are most central in a network according to “diffusion centrality” (a measure of network centrality which nests existing ones, and predicts the extent to which piece of information seeded to a network member diffuses in finite time). Using rich network data from 35 Indian villages, we find that respondents accurately nominate those who are diffusion central – not just traditional leaders or those with many friends. In a subsequent randomized field experiment in 213 villages, we track the diffusion of a piece of information initially given to a small number of “seeds” in each community. Seeds who are nominated by others lead to a near tripling of the spread of information relative to randomly chosen seeds. Diffusion centrality accounts for some, but not all, of the extra diffusion from these nominated seeds compared to other seeds (including those with high social status) in our experiment.
External Link(s)
Registration Citation
Citation
Banerjee, Abhijit et al. 2016. "Gossip: Identifying Central Individuals in a Social Network." AEA RCT Registry. November 27. https://www.socialscienceregistry.org/trials/1770/history/12075
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
Interventions
Intervention(s)
In this project, the researchers developed a model of information diffusion in which community members are able to accurately nominate the individuals most suited to disseminate information to the community ("gossip nodes"). By counting of how often they hear someone associated with a piece of information, people learn the correct ranking of their community members as information sources. The researchers term this ability to send information "diffusion centrality". Using network data from 35 Indian villages in Karnataka, the researchers find that diffusion centrality is the best predictor of whether an individual is nominated as a gossip node. The researchers tested their model by conducting an RCT where they track the diffusion of information in 213 different villages in Karnataka. They found that information diffusion was nearly three times larger from the nominated households as from randomly selected households, and that diffusion centrality accounted for much, but not all, of the difference.
Intervention Start Date
2015-02-05
Intervention End Date
2015-04-15
Primary Outcomes
Primary Outcomes (end points)
Number of calls to enter raffle
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
The researchers conducted a randomized field experiment to test whether information dispersion differs when it is seeded with gossip nodes instead of randomly chosen households. The information was that anyone who called a certain number was entered into a raffle for a free cell phone or cash prizes. The chance to win was non-rivalrous, and the call was free.

The field experiment included three treatment arms, each of which was administered in a different set of 71 villages in Karnataka. In the first treatment arm, the information was seeded with 3 or 5 randomly selected households. In the second treatment arm, the information was seeded with 3 or 5 households which include an individual with "elder status". In the third treatment arm, the information was seeded with 3 or 5 households that include a gossip node. Information dispersion was measured by the number of calls received from each village. More calls indicated that news of the raffle had spread to more individuals.
Experimental Design Details
Randomization Method
Randomization performed by computer
Randomization Unit
Villages were randomly selected into the three treatment arms; the households which were seeded with information were randomly selected conditional on their treatment assignment (i.e. from among all households in first arm, from among households with "elders" in second arm, and from households with nominated gossip nodes)
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
213 villages
Sample size: planned number of observations
213 villages
Sample size (or number of clusters) by treatment arms
71 villages - 3 or 5 random households; 71 villages information - 3 or 5 village "elders"; 71 villages - 3 or 5 nominated individuals
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Committee on the Use of Humans as Experimental Subjects at MIT
IRB Approval Date
2014-10-16
IRB Approval Number
1010004040
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is data collection complete?
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers